Master-slave Parallel Research Articles

An effective parallel approach for genetic-fuzzy data mining

Data mining is most commonly used in attempts to induce association rules from transaction data. In the past, we used the fuzzy and GA concepts to discover both useful fuzzy association rules and suitable membership functions from quantitative values. The evaluation for fitness values was, however, quite time-consuming. Due to dramatic increases in available computing power and concomitant decreases in computing costs over the last decade, learning or mining by applying parallel processing techniques has become a feasible way to overcome the slow-learning problem. In this paper, we thus propose a parallel genetic-fuzzy mining algorithm based on the master–slave architecture to extract both association rules and membership functions from quantitative transactions. The master processor uses a single population as a simple genetic algorithm does, and distributes the tasks of fitness evaluation to slave processors. The evolutionary processes, such as crossover, mutation and production are performed by the master processor. It is very natural and efficient to run the proposed algorithm on the master–slave architecture. The time complexities for both sequential and parallel genetic-fuzzy mining algorithms have also been analyzed, with results showing the good effect of the proposed one. When the number of generations is large, the speed-up can be nearly linear. The experimental results also show this point. Applying the master–slave parallel architecture to speed up the genetic-fuzzy data mining algorithm is thus a feasible way to overcome the low-speed fitness evaluation problem of the original algorithm.

Highly Scalable, Parallel and Distributed Adaboost Algorithm Using Light Weight Threads and Web Services on A Network of Multi-Core Machines

AdaBoost is an important algorithm in machine learning and is being widely used in object detection. AdaBoost works by iteratively selecting the best amongst weak classifiers, and then combines several weak classifiers to obtain a strong classifier. Even though AdaBoost has proven to be very effective, its learning execution time can be quite large depending upon the application e.g., in face detection, the learning time can be several days. Due to its increasing use in computer vision applications, the learning time needs to be drastically reduced so that an adaptive near real time object detection system can be incorporated. In this paper, we develop a hybrid parallel and distributed AdaBoost algorithm that exploits the multiple cores in a CPU via light weight threads, and also uses multiple machines via a web service software architecture to achieve high scalability. We present a novel hierarchical web services based distributed architecture and achieve nearly linear speedup up to the number of processors available to us. In comparison with the previously published work, which used a single level master-slave parallel and distributed implementation [1] and only achieved a speedup of 2.66 on four nodes, we achieve a speedup of 95.1 on 31 workstations each having a quad-core processor, resulting in a learning time of only 4.8 seconds per feature.

A Study on Parallel Computation for 3D Magneto‐Telluric Modeling Using the Staggered‐Grid Finite Difference Method

AbstractComputation time and memory requirements are two common problems for magnetotelluric (MT) modeling of three‐dimensional conductivity structure. We develop a new parallel processing scheme that can efficiently improve the computational speed of 3D MT modeling. The scheme of 3D MT modeling based on the staggered‐grid finite difference method is implemented in the frequency domain, and the calculation process of the EM field for each frequency is independent. Therefore, considering the naturally parallelizable character, the whole computation task of all frequencies can be divided into many minor calculation tasks for single or multiple frequencies, which will be assigned to different computing nodes and calculated in a parallel manner. In this work, by adopting master‐slave parallel mode and parallel computation with frequencies scheme, we have implemented the parallel computation of 3D MT modeling using MPI on the Dawn TC5000A high‐performance parallel platform. Furthermore, we tested our parallel algorithm of 3D MT modeling using two 3D theoretical models and analyzed the calculation efficiency on a multiple‐nodes computer, and the results show that the parallel algorithm is effective and efficient, which lays a solid foundation for subsequent three‐dimensional parallel MT inversion.

Parallel quantum-behaved particle swarm optimization

Quantum-behaved particle swarm optimization (QPSO), like other population-based algorithms, is intrinsically parallel. The master–slave (synchronous and asynchronous) and static subpopulation parallel QPSO models are investigated and applied to solve the inverse heat conduction problem of identifying the unknown boundary shape. The performance of all these parallel models is compared. The synchronous parallel QPSO can obtain better solutions, while the asynchronous parallel QPSO converges fast without idle waiting. The scalability of the static subpopulation parallel QPSO is not as good as the master–slave parallel model.

Development and application of a master-slave parallel hybrid multi-objective evolutionary algorithm for groundwater remediation design

Two primary goals of a multi-objective evolutionary algorithm (MOEA) for solving multi-objective optimization problems are to find as many nondominated solutions as possible toward the true Pareto front and to maintain diversity of Pareto-optimal solutions along the tradeoff curves. However, few MOEAs can achieve these two goals concurrently. This study presents a new hybrid MOEA, the niched Pareto tabu search combined with a genetic algorithm (NPTSGA), in which the global search ability of niched Pareto tabu search (NPTS) is improved by the diversification of candidate solutions that arose from the evolving population of nondominated sorting genetic algorithm-II (NSGA-II). The NPTSGA coupled with a flow and transport model is developed for multi-objective optimal design of groundwater remediation systems. The proposed methodology is then applied to a large field-scale groundwater remediation system for cleanup of large trichloroethylene plume at the Massachusetts Military Reservation in Cape Cod, Massachusetts. Furthermore, a master-slave (MS) parallelization scheme based on the Message Passing Interface is incorporated into the NPTSGA to implement objective function evaluations in a distributed processor environment, which can greatly improve the efficiency of the NPTSGA in finding Pareto-optimal solutions to the real-world applications. This study shows that the MS parallel NPTSGA in comparison with the original NPTS and NSGA-II can balance the tradeoff between the diversity and optimality of solutions during the search process and is an efficient and effective tool for optimizing the multi-objective design of groundwater remediation systems under complicated hydrogeologic conditions.

Parallel Processing and Performance Optimization of Meteorological Satellite Mass-Data Program

We implemented parallel processing of polar-orbiting meteorological satellite MERSI data aerosol retrieval program. The parallel implementation was based on Linux cluster architecture. The master-slave parallel programming mode on MPI parallel environment was applied. Performance optimizations are made in load balance, communication overhead, storage and system I/O according to the specific environment. In addition, parallel speed-up ratio and efficiency were analyzed to evaluate the experiment results. Experimental results demonstrate that the parallel techniques and the performance optimization methods proposed here can significantly improve the efficiency of satellite mass-data processing program.

Comparison of Multi-area Reactive Power Optimization Parallel Algorithm Based on Ward and REI Equivalent

By the means of introducing Ward and REI equivalent partition, the solution thoughts of multi-area power system reactive power optimization parallel algorithms based on Ward and REI equivalent are analyzed and explained in this paper, and the calculation methods of external equivalent network correction and outer coordination which used in two kinds of equivalent parallel algorithm are summarized, the multi-area reactive power optimization parallel calculation are implemented by using Matlab parallel computing platform and COW/Cluster hardware platform based on master-slave parallel calculation mode. Finally testing results on IEEE 39-bus, a 538-bus practical system and a 695-bus practical systems are used to analyze and compare the computing properties of two kinds of equivalent algorithm comprehensively in comparison with the centralized optimization method.

A Novel Spatial Interpolation Parallel Algorithm based on Genetic Expression Programming

Because of the limited environmental monitoring stations, the spatial data we have obtained by monitoring is local, discrete and limited. Therefore, speculating the environmental pollution parameters without monitoring station by using known monitoring data for spatial interpolation has become a hot research topic. In our preliminary work, an approach combining Genetic Expression Programming (GEP) evolution modeling with Delaunay, is presented to achieve automatic spatial interpolation. However, as a kind of genetic, GEP has the problem of premature convergence. In this paper, a GEP combining parallel algorithm based on MPI is proposed. By adopting the master-slave coarseness parallel GEP algorithm for spatial interpolation and introducing the subgroups migrating policy based on ring topology, this proposed algorithm has realized the coarseness scalable parallel computing which can run in the processor within a certain size to raise the ability of jumping out of the local optimum thus can avoid the problem of premature convergence efficiently. The result of simulation experiment shows that, as to the normal dataset and abnormal dataset of SIC2004, the GEP parallel algorithm based on MPI has the same predictable accuracy as the serial algorithms do, and the execution time is much less than serial algorithm, which can better satisfy the real-time demand of monitoring the content of radioactive material.

Parallel computation for debonding process of externally FRP plated concrete

In this paper, the three dimensional Parallel Realistic Failure Process Analysis (<TEX>$RFPA^{3D}$</TEX>-Parallel) code based on micromechanical model is employed to investigate the bonding behavior in FRP sheet bonded to concrete in single shear test. In the model, the heterogeneity of brittle disordered material at a meso-scale was taken into consideration in order to realistically demonstrate the mechanical characteristics of FRP-to-concrete. Modified Mohr-coulomb strength criterion with tension cut-off, where a stressed element can damage in shear or in tension, was adopted and a stiffness degradation approach was used to simulate the initiation, propagation and growth of microcracks in the model. In addition, a Master-Slave parallel operation control technique was adopted to implement the parallel computation of a large numerical model. Parallel computational results of debonding of FRP-concrete visually reproduce the spatial and temporal debonding failure progression of microcracks in FRP sheet bonded to concrete, which agrees well with the existing testing results in laboratory. The numerical approach in this study provides a useful tool for enhancing our understanding of cracking and debonding failure process and mechanism of FRP-concrete and our ability to predict mechanical performance and reliability of these FRP sheet bonded to concrete structures.

A PARALLEL SOLVER FOR THE DESIGN OF OIL FILTERS

Nowadays, it is widely recognized that computer simulation plays a crucial role in designing oil filters used in the automotive industry. However, even a single direct simulation of the flow usually requires significant computational resources. Thus, it is obvious that solution of optimization problems is only feasible using parallel computers and algorithms.In this paper, we present a general master-slave parallel template, which was specially designed for the easy integration of direct parallel solvers into a parallel optimization tool. We show how an already existing direct solver for the 3D simulation of flow through the oil filter is integrated into our template to obtain a parallel optimization solver. Some capabilities and performance of this solver are demonstrated by solving geometry optimization problem of a filter element.

Genetic algorithm optimization and blending of composite laminates by locally reducing laminate thickness

Composite panel structure optimization is commonly decomposed into panel optimization subproblems, with specified local loads, resulting in manufacturing incompatibilities between adjacent panel designs. A new method proposed here for constructing globally blended panel designs uses a parallel decomposition antithetical to that of earlier work. Rather than performing concurrent panel genetic optimizations, a single genetic optimization is conducted for the entire structure with the parallelism solely within the fitness evaluations. A genetic algorithm approach, based on locally reducing a thick (guide) laminate, is introduced to exclusively generate and evaluate valid globally blended designs, utilizing a simple master–slave parallel implementation, implicitly reducing the size of the problem design space and increasing the quality of discovered local optima.

Hyper-Rectangle Selection and Distribution Algorithm for Parallel Adaptive Numerical Integration

In this paper we consider parallel numerical integration algorithms for multi-dimensional integrals. A new hyper-rectangle selection strategy is proposed for the implementation of globally adaptive parallel quadrature algorithms. The well known master-slave parallel algorithm prototype is used for the realization of the algorithm. Numerical results on the SP2 computer and on a cluster of workstations are reported. A test problem where the integrand function has a strong corner singularity is investigated. A modified parallel integration algorithm is proposed in which a list of subproblems is distributed among slave processors.

Scheduling Tasks in Master-Slave Parallel Processing Systems

The problem of deterministic scheduling of tasks in the master-slave environment is analyzed in this paper. It is assumed that the slave reads data from the master, processes it, and sends the results to the master. This paradigm can also model a wider class of distributed applications. The goal is to find the shortest schedule. The general problem is computationally hard. However, any greedy heuristic delivers solutions not worse than twice the optimum when tasks are preallocated, and three times the optimum when tasks are not preallocated. Polynomially solvable cases are considered.

Generalized speculative computation of parallel simulated annealing

Simulated annealing is known to be highly sequential due to dependences between iterations. While the conventional speculative computation with a binary tree has been found effective for parallel simulated annealing, its performance is limited to (logp)-fold speedup due to parallel execution of logp iterations onp processors. This report presents a new approach to parallel simulated annealing, calledgeneralized speculative computation (GSC). The GSC is synchronous, maintaining the same decision sequence as sequential simulated annealing. The use of two loop indices encoded in a single integer eliminates broadcasting of central data structure to all processors. The master-slave parallel programming paradigm simplifies controlling the activities ofp iterations which are executed in parallel onp processors. To verify the performance of GSC, we implemented 100-city to 500-city Traveling Salesman Problems on the AP1000 massively parallel multiprocessor. Execution results on the AP1000 demonstrate that the GSC approach can indeed be an effective method for parallel simulated annealing as it gave over 20-fold speedup on 100 processors.